Written by Asaf Somekh, CEO & Founder at iguazio

We are witnessing a technological revolution that is dramatically changing the way we live and work. The speed at which technological breakthroughs are occurring has no precedent in previous periods of transformation. This revolution is disrupting almost every industry in every country.

Most of the technological revolutions now taking place or about to take place are based on AI. Adoption of AI in businesses across the US, Europe and China has risen sharply over the past year, to 34 percent. AI technology uses algorithms to analyze large data sets, detect patterns, extract insights, and make decisions accordingly. Israel is widely celebrated as an AI powerhouse, despite its population size.

AI technologies make it possible to use the massive amounts of accumulated data and make use of them. The growing market around AutoML solutions has made data science accessible to a larger segment of organizations. However, according to industry analysts an estimated 85% of data science projects, which have shown great promise in the lab, never make it to production. This is due to the challenges of transforming an AI model, which is functional and shows great promise in lab conditions, to a fully operational AI application that can deliver business impact at scale and in real business environments.


The potential use cases for data science are truly exciting. But the elaborate challenges of operationalizing machine learning can--and often does--impede companies from bringing innovative solutions to market. Software development has by now become a repeatable efficient practice, but for the AI industry, the complexities involved in machine learning applications means there is still a lack of standards and widespread best practices.

This is changing. MLOps (Machine Learning Operations) is an emerging discipline that echoes DevOps practices for machine and deep learning. MLOps decreases time and effort by automating the tasks involved with deploying, monitoring and managing AI applications in production. As the MLOps field evolves, new technologies like feature stores are emerging to break down silos between data scientists, data engineers and DevOps practitioners, by allowing everyone on the team to build, share, reuse, analyze and monitor features in production. This unified approach to feature engineering accelerates the path from research to production and enables companies to develop, deploy and manage AI at scale with ease.  As companies across industries weave AI/ML applications into their processes, IT leaders must invest in MLOps to drive real business impact.

Operationalizing AI in Critical Care Environments

The ARC Innovation Center at Sheba Medical Center, ranked one of the top ten hospitals worldwide by Newsweek,  is a standout example of how AI can be operationalized to dramatically improve healthcare. Sheba Medical Center is the largest hospital in Israel and the MEA region, and possesses one of the world’s largest reservoirs of health data. ARC recently launched a new project that brings urgent real-time predictions to the ICU floor.  Harnessing data from various sources, such as real-time vital signs, x-rays and historic patient records, they use advanced machine learning algorithms to optimize patient care, predict COVID-19 patient deterioration and even control the flow of cars to parking spaces . Real-time dashboards surface alerts for doctors and prioritize patient intake, so the medical center can respond quickly and dramatically improve outcomes for all involved.

The massive strain placed on companies and health organizations by COVID-19 has only emphasized what we knew before: it is absolutely vital for businesses who want to survive the current situation to create a competitive advantage by bringing AI innovations to market quickly. Companies must adapt to a rapidly shifting environment by focusing on developing and deploying AI more efficiently, without the excessive costs and lengthy timeframes that might have seemed reasonable just a year ago. And now more than ever, monitoring AI applications for concept drift is critical, as human behavior changes dramatically from week to week during these unpredictable times, leaving AI models unusable due to a change in the very things they were built to predict.

With the new Israeli government setting a goal to increase the percentage of high tech employees from 10% to 15% of the overall workforce, AI technologies will be a critical growth engine for the Israeli economy. Israel’s past success establishing thought leadership and market dominance in the cybersecurity market bodes well for its ability to overcome the current obstacles it faces on the path to global AI leadership.  MLOps will be a big facilitator in this path, enabling more and more companies to see real business value from their AI endeavors, in a short timeframe and with a lean team.